The station is located on the summit (plateau) of the second highest point of the Hoggar mountain range in the Saharan desert. The site is very remote at a distance of 50 km from Tamanrasset. Touristic activities in the area are limited due to difficult access to a few dozen visitors per day. Vegetation is extremely sparse.
The Assekrem station allows us to get a clean read with little local bias due to the remoteness of the location. This allows us to get atmospheric readings of the specific gases as opposed to local short-lived and erratic gas emissions. For example, CO2 travels and mixes in the atmosphere due to its long lifetime, allowing us to see this value as opposed to having a station right next to an industrial city, where readings will largely depend on the production for that specific factory. The high position of the station can indicate that the readings are safe from biases induced from the thermal inversion layer of the atmosphere, including breeze, wind changes, and reduced or enhanced molecular transport. This provides a sturdy ground for atmospheric readings.
The data is collected in the Assekrem station in Algeria. Located at 23.2625° North, 5.6322° East, and at a 2710 meter elevation, this is one of the most remote stations and one of the only ones in the entire Saharan region. Many different techniques are used to monitor gases on Earth, but all have some type of limitations. For this specific case, the station uses in-situ measurements, which involve a single location using high quality, wide set of instrumentation, which can take precise measurements of a small and specific geographical point. two main limitations arise from this method:
In situ measurements provide what is best described as direct observations of the system. The great advantage is the way we can make use of a diverse set of instrumentation which can take the most accurate readings we can get.
Prophet is an algorithm developed by Facebook’s data team in order to automate a routine for forecasting trends given equally spaced data points (Taylor et. al, 2017). The strategy is to frame the forecasting problem as a curve-fitting by disentangling two components:
The parameters of the model are adjustable (eg., adding seasonalities, specifying forecast frequency, etc) which offers us the advantages of the Bayesian approach since the forecasts are derived from the posterior distribution.
from fbprophet import *
import pandas as pd
import plotly.offline as py
import plotly.graph_objs as go
def forecasting(gas_name, gas_code, months, unit):
gas_data = pd.read_table('C:/Users/Taha/Desktop/algeria_ghg/'\
+gas_code+'_ask_surface-flask_1_ccgg_month.txt',
sep='\s{1,}', names=['stations','year','month','y'], engine='python')
gas_data['ds'] = pd.to_datetime(gas_data[['year', 'month']].assign(DAY=1))
model = Prophet(weekly_seasonality=False, daily_seasonality=False)
model.set_auto_seasonalities
model.add_seasonality(name='monthly', period=30, fourier_order=5)
model.fit(gas_data)
future = model.make_future_dataframe(periods=months,freq='M')
forecast = model.predict(future)
py.init_notebook_mode()
fig1 = plot.plot_plotly(model, forecast)
fig1.update_layout(title=str(gas_name)+' Forecast', xaxis_title='Time',
yaxis_title='Parts per '+unit+'illion (PP'+unit+')')
py.iplot(fig1)
fig2 = plot.plot_components_plotly(model, forecast)
fig2.update_layout(title=str(gas_name)+'<br>Trend & Seasonality', height=500)
py.iplot(fig2)
'''
Measurements are reported in units of
micromol/mol (10^-6 mol CO2 per mol of dry air or parts per
million (ppm)). Measurements are directly traceable to the
WMO X2007 CO2 mole fraction scale.
'''
forecasting('Carbon Dioxide', 'co2', 240, 'M')
'''
Measurements are reported in units of nanomol/mol
(10^-9 mol CH4 per mol of dry air (nmol/mol) or parts per billion
(ppb)) relative to the NOAA 2004A CH4 standard scale.
'''
forecasting('Methane', 'ch4', 240, 'B')
'''
Carbon monoxide mixing ratios in these files are reported
in units of nmol/mol (10^-9 mole CO per mole of dry air
or as part per billion by mole fraction (ppb)) relative
to the NOAA/WMO CO scale (Novelli et al., 1991, Novelli
et al., 2003).
'''
forecasting('Carbon Monoxide', 'co', 240, 'B')
'''
N2O measurements are reported in units of nanomol/mol (10^-9 mol N2O
per mol of dry air (nmol/mol) or parts per billion (ppb)) relative
to the NOAA 2006A N2O standard scale.
'''
forecasting('Nitrous Oxide', 'n2o', 240, 'B')
'''
SF6 measurements are reported in units of picomol/mol (10^-12 mol
SF6 per mol of dry air (pmol/mol) or parts per trillion (ppt))
relative to the NOAA 2014 SF6 standard scale.
'''
forecasting('Nitrous Oxide', 'sf6', 240, 'T')
Dlugokencky, E.J., J.W. Mund, A.M. Crotwell, M.J. Crotwell, and K.W. Thoning (2019), Atmospheric Carbon Dioxide Dry Air Mole Fractions from the NOAA ESRL Carbon Cycle Cooperative Global Air Sampling Network, 1968-2018, Version: 2019-07, https://doi.org/10.15138/wkgj-f215
Taylor, S.J., & Letham, B. (2017). Forecasting at Scale. PeerJ Prepr., 5, e3190. https://www.semanticscholar.org/paper/Forecasting-at-Scale-Taylor-Letham/ab1f816ce79817a09487ea7866c95ce930d37497